高级人工智能

摘要

国立台湾大学于天立教授的视频笔记

Search

For example:

Inital state :No queen on the board

Actions : Add a queen on the board where the square is empty

Transition model : return the board with a queen added to the spacified squere

Goal test : 8 queens are on the board ,none attacked

Path cost : numbers of trials

Different States space will be have different Solution space

b:maximum branching factor of the search tree

d:depth of the least-cost solution

m:maximum depth of the state space (may be $\infty$)

Uniformed Search 无信息搜索

Uniformed strategies use only the information available in the problem definition

Breadth-first search(BFS)

1,Expand the shallowest unexpanded node

2,FIFO queue

Completeness	Optimality	Time complexity	Space complexity
yes(if b is finite)	yes	$O(b^d)$	$O(b^d)$

Uniform-cost search

an special example of BFS

expand the unexpanded node with the lowest path cost

Priority queue orderd by $g(n)$

equivalent to BFS if step costs all equal

For TREE-SEARCH , priority queue gives the cheapest path first

For GRAPH_SEARCH , if the node is already in the frontier , need to find the minimum cost , and call DECREASEKEY as needed

Completeness	Optimality	Time complexity	Space complexity
yes	yes	$O(b^{1+\lfloor \frac{C^{\ast }}{ϵ}\rfloor })$	$O(b^{1+\lfloor \frac{C^{\ast }}{ϵ}\rfloor })$

Depth-first search(DFS)

1,expand the deepest unexpand node

2,stack

Completeness	Optimality	Time complexity	Space complexity
No	No	$O(b^m)$	$O(m)$ !!! linear

Depth-limited search

Iterative deepening search(IDS)

Call DLS iteratively with increasing depth limit

Seems to be wasteful , but actually not

Combine the benefits of BFS and DFS

for depth = 0 to $\infty$
    result = DLS(problem,depth)
    if result != cutoff
        return result

Completeness	Optimality	Time complexity	Space complexity
Yes	Yes only if step cost = 1	$O(b^d)$	$O(bd)$

Bidirectional Search

Summay

a:if b is finite

b:if b is finite and step cost >= $ϵ$

c:unless $,$ >= d

d:if b is finite and both direction use complete search like BFS

e:if all steps cost are identical

Criterion	Breadth-first search	Uniform-Cost	Depth-first search(DFS)	Depth-limited search	Iterative deepening search(IDS)	Bi-Directional
Completeness	Y a Y^a Y